A social network represents social relationships between individuals, groups, or organizations and is a classic example of a complex network. Social network analysis focuses on understanding the connections between people or entities and analyzing the influence of behavior, communication, and information flow between entities. Link prediction, a fundamental research domain in social network analysis, aims to predict the formation of new connections between individuals or entities. Conventional link prediction techniques, such as the Jaccard coefficient, preferential attachment, and common neighbors, rely on heuristics. While these methods are computationally simple, they frequently fall short of capturing the complex, non-linear patterns found in real-world networks. These approaches are also limited, particularly when the networks are sparse, big, or continuously changing. The modern deep learning approaches overcome these drawbacks by extracting topological and node attributes. In this article, we utilize the five deep learning models that capture the non-linear dependencies that exist in the real-world network that are crucial for link prediction. Through these models, the higher-order interactions are captured that are otherwise left unnoticed during the future link prediction techniques.